During last weeks’ GTC Asia conference, NVIDIA announced that the company is opening up CUDA standard by sharing the source code of the new CUDA LLVM-based compiler.

For as long as we know, NVIDIA kept CUDA (Compute Unified Device Architecture) proprietary in order to keep a tight grip between CUDA software and CUDA hardware. When you compare such approach to for example AMD or Intel that had various limitations in their implementations of OpenCL (NVIDIA is also considered to have one of the best, if not the best OpenCL app support).

However, the company realized that with Fermi and upcoming Kepler architecture, the compute capabilities have advanced to that state that there is no other way but to open the compiler code to various academic institutions and qualified software vendors.

According to NVIDIA, "LLVM is a widely-used open source compiler infrastructure with a modular design that makes it easy to add support for new programming languages and processor architectures.  It is used for a range of programming requirements by many leading companies, including Adobe, Apple, Cray, Electronic Arts, and others.
The new LLVM-based CUDA compiler, which is enhanced with architecture support for NVIDIA?s parallel GPUs, is included in the latest release of the CUDA Toolkit (v4.1), now available to the public."

You can register here: http://developer.nvidia.com/cuda-source.